4.4* COULD IT BE SUBVERTED?

To evaluate the prospects of a country that adopts the proposed constitution, we must also go beyond the foregoing detailed discussions of incentives and the claims of Chapter 2. We must consider the question, if an organized interest group wants to subvert the clear purpose and structure of this constitution, short of a military coup could they do it? In other words, in practice would this constitution be self-enforcing?

In addressing this question, we must draw the distinction between minor, occasional failures and a major, decisive breakdown of the constitutional plan. Every government has occasional instances of corruption and other offenses by officials, just as fraud, embezzlement, self-dealing, and other such offenses occur in the private sector. There is no reason to expect such problems to be either more or less frequent under the proposed constitution, except to the extent that it successfully provides for better monitoring of official performance than do existing constitutions and laws. It is fair to claim, I think, that the opportunity our plan provides to a second-best official legislature to recoup some bonus payments that it fails to win in the referendum, through its role in supervising the auditing and inspectorate function, should provide an effective level of monitoring of official wrong-doing. If so, it will match or better the performance of the best now observed anywhere in this official function, taking due account of cultural differences from one country to another in such matters.

By contrast, a major failure would be a successful effort to induce the government to spend money wastefully to benefit an organized group of voters at the expense of other voters, followed by another similar successful effort, and so on, so that the intended structure set out in the constitution breaks down. We must assume that, however clear the language of the constitution may be on this point, if any crack in its armor leaves an attractive incentive to organized groups to attempt such an effort, they will attempt it. Is there a crack in the armor such that they could succeed?

The protective barriers that the draft constitution constructs around official legislatures make it unlikely that an interest group would attempt to bribe legislators. There is no legal way to find their identities and to approach them. Moreover, to limit the agenda before the voters, the interest group would have to bribe all official legislatures. Then private legislatures would step in with sound counterproposals, if anything were appropriate besides the constitutionally mandated ‘status quo’ budget. To keep out sound alternative proposals and to control the official assessment of these proposals, the interest group would have to bribe the Election Commission also. It strains credulity to suppose that there is a profitable way to bribe so many people enough to overcome their legal incentives. Moreover, if someone did there would still be the remedy of lawsuits. Under the proposed constitution, this remedy would apply equally to public sector programs and to private-sector goods and services. Laws relating to false advertising and fraud would be the same for both sectors.

Instead, what if an interest group were to advance a private proposal, using the mechanism authorized by the constitution, and use emotional arguments, etc., to sell it? To make this approach work, the group would also have to win referendum approval of phony measures of success, to avoid prohibitive penalties on performance. This plan would have to run the gauntlet of authoritative review in official legislatures, and of the sound arguments pro and con that these legislatures would produce for the referendum. If someone succeeded, would a subsequent effort at similar deceit succeed equally well? How often do product misrepresentations succeed in the private sector? When they do succeed, how materially do they subvert welfare and reduce general efficiency, compared to any known alternative approach to resource allocation? Although one can surely find examples where private businesses have sold worthless or overpriced products, it is hard to find any that compare with the excesses and misrepresentations by existing governments.

EXAMPLE(1)

Consider the following example of deceit by governments. We noted in Chapter 1 the absurdly generous system of welfare payments (including Medicaid) in the U.S., the costly Medicare program for the elderly, and the perverse incentive effects of the social security program. What we did not point out there is that the cumulative effect of these programs, if continued on their present basis with its future projections, imply impossibly high tax obligations for future generations of taxpayers. That is, this system of transfers is a pyramid game that will cheat young people who still have several decades to work before their retirement (Kotlikoff, 1992). For more than a decade, the Trustees of the Social Security Administration have published estimates using straightforward accounting methods showing that future tax increases for social security alone are improbably high. Politicians of every political persuasion know that the combination of transfer programs is unsustainable in the long run. Under either political party’s proposals, however, the game will continue until it breaks down or unless, prior to that, there is a substantial correction.

Rather than assert a bald claim that nothing that bad would happen under the proposed constitution, let’s examine the particulars. Suppose that an organized group purporting to represent the elderly mounts a campaign to increase Medicare and social security benefits to elderly persons, perhaps working in concert with another group purporting to represent the poor, who would propose increased supplements for this or that dire need of the poor. The constitutional rule for the referendum requires that the legislative process shall generate, for voters’ use, a full and fair account of the waste occasioned by these programs and of the eventual collapse of the pyramid game the increased benefits would create. Most voters in the U.S. do not now know about the present pyramid game, because no one has been willing to spend the required sums to inform them, and no politician has an incentive to tell them. If they knew, it is hard to believe that it would go on. Under the proposed constitution, the official legislatures would have the relevant facts, would report them, and would indicate that the proposals for increased benefits had no chance of passage. To prevent this full and fair account from reaching the voters the organized groups would have to bribe the legislatures and the Election Commission handsomely enough to offset any reduction in prospective bonus payments and compensate for the enormous risks involved. The bribes would be paid to suppress the relevant facts and to substitute the kinds of emotional arguments we now see in politics. That’s a lot of money and a large number of persons to bribe.

Moreover, there is a large group who would be harmed, namely taxpaying people from twenty to about fifty years old, who would pay much of the cost of the benefits to the present elderly and then would receive sharply reduced benefits themselves upon their retirement. Under the rational legal structure encouraged by the constitution, they could set up a class-action suit and win. They would win in the initial court proceeding because the court would be impartial and all the facts would come out. With so much bribe money going to so many people, it would be easy to prove. Even if bribery could not be proved, malfeasance could. In the final appeal to an official legislature (perhaps necessarily a reconvened previous one), the offended groups would be fully represented, the harm to them would outweigh the benefits to the elderly and poor, because of the inefficiency of these programs. Hence under the VCG mechanism the suit would have a high probability of winning at the top level. The risk of this outcome would prevent the acceptance of bribes and therefore would prevent the formation of the organized interest groups.

To judge the likelihood of all this we should bear in mind that existing governments have two features that governments under our proposed constitution would not have. First, it is taken for granted that their programs benefit organized groups at the expense of other citizens. This constitution expressly prohibits and prevents income transfers through such programs except for constitutionally specified redistribution to the poor. Second, they have many channels for legalized or tolerated bribery, such as campaign contributions, employment on favorable terms after leaving government service, vacation trips and other gifts, and so on. This constitution almost eliminates campaigns and provides for clear prohibitions of bribery by other means. Large-scale bribery would be both risky and unprofitable.

Rather than resort to bribery, however, a group might resort to conspiracy to falsify declarations of willingness to pay on an issue that would in any case appear in a referendum. Although we show in Chapter 5 that such conspiracies will fail in the case of items dominated by Thompson insurance, we need to address here the case of items that are wholly or largely uninsurable, in the sense that because of income effects households will not insure against unfavorable outcomes on these items. When we look closely, what we find is that although there is a risk in this case, it is not serious.

EXAMPLE(2)

Besides providing for criminal sanctions against conspiracy to falsify declared harms in referenda, the constitution’s internal checks largely eliminate the potential rewards from such conspiracies. Consider the following example. A large country of 100 million voting families, with median family incomes of $50,000, has a defense budget of $100 billion, which comes to $1000 per family. No family buys Thompson insurance against an unwanted defense budget outcome, so that the budget results exclusively from declarations of uninsured harm. Half the families are ‘hawks’, who want a budget of $200 billion, and half are ‘doves’, who want a budget of $50 billion. Hawks and doves are scattered among the population evenly, so that there is no way to identify them by demographic groups. At the equilibrium budget, a ‘hawk’ family’s willingness to pay is $15 per billion of additional budget, of which the family’s tax share would be $10, leaving a net excess willingness to pay of $5. A dove family’s willingness to pay is $5 per billion of additional budget, as against the tax share of $10. In both cases the elasticity of each willingness to pay schedule, as a function of budget size, is about – 2. The VCG tax for each family at the equilibrium budget works out at $.25x10^-7, a ridiculously small fraction of a cent each.

In this circumstance, Margolis (1982a) argues that, even without collusion, voters will naturally tend to overstate their willingness to pay to obtain budget changes. If the hawks and doves all follow Margolis’s advice to ‘spend a penny’ of VCG tax to move the budget in their respective preferred directions, their misrepresentations will cancel and the budget will remain the same. Suppose, however, that the hawks alone have a reckless tendency or a conspiracy to exaggerate in this way, whereas the doves all tell the truth. To raise a hawkish family’s VCG taxes to one penny, they would multiply by a factor of 400 their $5 net excess willingness to pay for the first $1 billion of defense spending, raising it to $2,000, and reduce their elasticity of demand by a corresponding factor so as to have their net willingness to pay to fall to zero at $200 billion, as before (or to a figure just above that.) The conspiracy would succeed in raising the defense budget to $200 billion, exactly or approximately, if all hawks behaved alike, because at all lower budget figures their excess willingness to pay in the aggregate would swamp the negative willingness to pay of the doves.

The constitution would permit all this to happen only if it had no internal checks on the reasonableness of the declared schedules of willingness to pay. The hypothetical exaggeration would imply that each hawkish family would be willing to pay, if necessary, about $700,000 to get the desired defense budget. The problem with that is that the typical hawkish family doesn’t have that much money, so that the declared willingness to pay is inconsistent with budget-constrained rational preferences. The constitution’s requirement in that respect, in Article II, Subsection 10(b), requires the voter to state credible willingness to pay figures, whose implied effect shall not exceed a reasonable figure such as 10 percent of the voter’s income. This restriction would allow the hawks to exaggerate their net excess willingness to pay only about 30-fold, from $5 per family to $150 per family. With that restriction, the effect of the conspiracy on the defense budget would be less than $1 billion, which is hardly worth risking criminal sanctions in the case of actual conspiracy. Far fewer than 100 percent of hawkish families would take this risk, as a rule, so that such conspiracies or quasi-conspiratorial behavior would have very little effect in practice.

We can therefore conclude that neither bribery nor conspiracy would be a significant threat to the basic plan of the constitution. We examined this issue from another perspective in the appendix to Chapter 2.

At the same time, these examples and arguments cannot be claimed to rule out any possibility of fads, fashions, and temporary surges of public passion. These occur in the private sector, and have in one way or another (except perhaps in ancient Lokris, as noted in Chapter 3) always been a hallmark of government of every kind. (This point ties into our discussion of common law in Chapters 2 and 3 and of the herd instinct in the appendix to Chapter 2.) We must nevertheless recognize that the features of present-day representative governments that we reviewed in Chapter 1, which invite irresponsible behavior by almost everyone, magnify and exacerbate this problem. By requiring or approximating unanimous consent on every public issue, expressed in budget-constrained dollar terms, this constitution provides the most reliable legal barrier against hasty and foolish policies that one can design, short of a method like that used in ancient Lokris. It would of course be helpful if there were also a natural public conservatism to back it up, and to the extent that a country has this attribute our constitutional design will protect this public against manipulation. Where a country lacks it, there is no institution that will protect it from its own foolishness.

4.5* IS IT OVERLY SPECIFIC?

A further question is, have we written too many details into this draft constitution? The draft constitution includes some particulars that may be gratuitous, such as the requirement to minimize bureaucracy and contract out everything possible to the private sector. The rational incentives provided by the legislative process should be enough to guarantee such outsourcing without a specific constitutional mandate. An opposing view is that the outsourcing provisions are so fundamental a part of the objective to avoid government waste, and that bureaucracies can so easily control the flow of information relating to their efficiency, that there needs to be a constitutional barrier to overgrowth of government employment. This barrier could still be overcome, but only with much greater difficulty. If the leaders of a new country were supremely confident that this language is gratuitous, they could of course omit it from their constitution.

Indeed one can think of many examples of good law and good practice that one might want to entrench in the constitution, and it is important to resist the temptation to overdo it. Things that can be safely left to be worked out later by the due process specified in the constitution should in general be left out, so that the constitutional decision focuses only on its essentials. Its likelihood of acceptance is low enough already, without excess baggage. Conceivably that could be true of the provision about outsourcing.

However, (unlike Mueller) we are not discussing a campaign for a constitutional convention in a country that is already a going concern (Mueller, 1996, Ch. 22). Every such country already has too many entrenched interests for such a convention to have an appreciable chance of success. The only hope for its adoption anywhere is in a new country of unusually prudent voters with unusually perspicacious leaders. I haven’t noticed any such country lately, but we should be properly prepared for that eventuality by having a sound plan ready to offer them. For that eventuality, the provisions on outsourcing might be a good selling point rather than an impediment to adoption.

Apart from the language on outsourcing, the reader may be startled to notice that the constitution is silent about the administration of tax collection, and in particular avoids any mention of it in the list of career government employees. A new government could of course change this feature if its leaders thought that appropriate. I would argue, however, that there is no reason to do so.

The tax rules under this constitution require each household to report its demographic characteristics identifying its population subgroup memberships and its property. The household would then receive, with its information package for a referendum, its tax implications for each program group in the referendum. The Election Commission would contract with private firms to provide a computer system for translating the voting outcome into household Lindahl taxes, Thompson insurance net balance payable or receivable, and VCG taxes. The private firms would post bond and would be subject to inspection and audit to protect the confidentiality of household information. Private auditors could provide the service of verifying household ‘tax returns.’ In a country that has reputable auditing firms, their reputations would generally provide the main guarantee of the integrity of their work. In a country that does not have such firms, the Election Commission might use specially trained government employees in a temporary role until such private firms could establish reputations and assume this function. (When they did, it would be appropriate to allow such firms to hire the government employees being dismissed, provided that these had no role in the selection of the firm that wins the government contract.) Although this plan differs markedly from past practice (except for tax-farming in eighteenth Century France), I see no reason to believe that official tax collectors have an advantage that would outweigh the disadvantage of giving them a monopoly of their function.

Problems of abuse arise not from the use of private or public auditors and tax collectors but rather from a lack of clarity in a household’s legal obligation. Where the legal obligation is clear it is comparatively easy to curb corruption and other abuses. The main difficulty arises in the assessment of property taxes, performed by the official legislatures and the Election Commission, because of possible ambiguities in the classification of items of property. A special set of tax courts might be an efficient part of the system to deal with tax assessment problems, as we now have in the U.S. Such courts would have the same ground rules for competition and tenure as the rest of the judiciary under Article V.

Therefore I conclude that the sound rule is to have tax collection and auditing contracted out to private firms just as is the case for almost everything else the government does under this constitution.

Many readers will understandably feel a sense of outrage at a proposal whose effect is to institutionalize the influence of wealthy persons in a country’s policy. In modern countries with governments chosen by election, nearly everyone’s sense of fairness suggests that every person’s vote should count the same as everyone else’s, among those eligible to vote. Many of those who hold this view almost to the point of religious devotion have noticed that the practice and the results of representative democracy in the modern world leaves something to be desired, and their reaction, without serious or credible analysis of free-rider problems, interest groups, and so on, is to suppose that it can be fixed.

Those of us who doubt that there is a way to fix the egregious problems of existing governments, understanding the obstacles more or less well, have generally been resigned to the consequences. Without undue admiration for or devotion to the workings of what we call representative democracy, we have had to accept Churchill’s observation that every known way of governing works out worse. The point of this book is that his statement no longer applies: there is a better way.

It is enough to state and support this claim without overly belaboring it. A country using the constitution proposed here could achieve the high levels of output and growth that we indicated in Chapter 1. How much economic welfare would sincere advocates of traditional democracy be willing for their countries to sacrifice for that supposed ideal? Over time, the material sacrifice would be enormous. Moreover, the sacrifice would be pointless. If we think of government purely and simply as a provider of services, why shouldn’t its provision of services be handled under procedures that give the same kinds of results as do private markets?

To see this point, we should look at some examples of how the system would work, from the voter’s point of view. In Section 4.6 we consider several cases covering a broad spectrum of issues, showing in general terms how the decision mechanism in the proposed constitution would resolve them. Then in Section 4.7 we examine the role of the voting family, and the incentives it faces, with referenda on a ‘binary’ (yes-or-no) decision and on the choice of a point within a continuous range.

4.6* THE PAYOFF IN MORE CONCRETE TERMS

Chapter 2 mentioned specific instances such as import controls for sugar that would vanish under the draft constitution proposed here, and others that would undergo drastic modification. By and large, however, our discussion has focused on ideas and principles, and can benefit now from further examples.

EXAMPLES(3-6)

The following examples illustrate the operation of our proposed rules. On the general principle involved, consider first the following example. A group of wealthy persons, living in an expensive subdivision, desires the construction of a yacht harbor suitable for their yachts as well as for smaller craft. They would be willing to pay a substantial part of its cost of construction, but the project can be economic only if other users pay part of the cost. A proposal on a referendum ballot to build a multi-purpose facility, under the proposed constitution, would lead them to declare valuations and accept Lindahl taxes for this improvement that were more than proportionate to their wealth, while other potential users would have comparatively small tax shares and would declare low valuations reflecting their more modest prospective use of the harbor. The proposed plan would have a sound marginal-cost based set of user fees that would also bear heaviest on the owners of large yachts. The wealthy yacht owners would thus pay most of the cost, and would have a correspondingly high influence on the approved construction budget. That seems entirely reasonable to me. Everything the government does, except the constitutionally mandated set of taxes to finance public assistance to destitute persons, would have tax financing based on the benefit principle, which would imply high influence for wealthy persons only in the case of projects of particular interest to them.

Next, we should look at some larger-scale issues that might come up in a national legislature. First, consider a project to dam a river, in a country with many such opportunities, to provide irrigation, flood control, hydro-electricity, and recreation. Second, assuming that there is no other special procedure for this decision, consider a proposal to abolish a symbolic, powerless, monarchy and to retire members of the royal family to the status of ordinary citizens. Third, consider the choice between a military draft and an all-volunteer armed force.

In the first example, a multipurpose dam, assume that, besides its financial consequences, its only effects are direct amenities that compete with other recreational opportunities of easily measured market value. That is, it does not reduce the habitat of an endangered species, nor destroy any special scenic area or archaeological treasures. The land it uses can be appraised accurately at market prices. Its irrigation water will serve known farmlands whose additional output will not disturb relative prices, nor do the project’s other goods and services, such as electricity. In short, assume that customers who do not receive these benefits nor supply resources to this project have no interest in whether it is built.

In this case the decision process in the legislature would reveal zero interest in the project outside the region that it serves. The assumption of no effect on relative prices of farm products and electricity implies that there would be no consumer interest in the project, except for those who expect recreational benefits and flood protection. The legislative decision process would reveal the farm demand schedule for the irrigation water, and the value that local households place on recreational and flood control benefits. The value of the local electricity would be measured by its market price. The estimated Lindahl taxes for the project, which in this case would be quite accurate, would fall on the following groups: (a) farmers, to collect their gain in land values, if any, after their prospective payments for water; (b) those property owners who have a rise in their property values due to flood protection; and (c) those property owners whose property values rise due to increased actual or imputed rents due to the recreational benefits.

Suppose that the legislative result indicates that the estimated Lindahl taxes thus identified, plus the revenues from the sale of electricity, water, and admissions to the recreational facilities, would pay for the dam at the appropriate rate of return to capital. When the issue is put to a referendum, if the indication is correct, some voters would buy Thompson insurance, and none other than gamblers and those who disagree with the odds given by the insurance premiums would declare values under the VCG part of the mechanism. If in fact the benefits to the affected households exceeded the costs, the referendum would approve the project, and not otherwise. Apart from gamblers, persons who disputed the official odds, and random errors in the matching of taxes to beneficiaries, the winning side would receive unanimous support. The outcome might nevertheless be hard to predict, with odds not far from even, if the project had benefits approximately equal to costs.

The second example differs sharply from the first in that it involves direct utility of a symbolic activity about which people may feel strongly, and for which there is no market counterpart. To emphasize this point, assume that the royal family covers all its expenses from its own wealth, which will be unaffected by the decision. Further, assume that the royal family’s tastes for and uses of economic goods will be unaffected by the outcome, and that the tourist trade, if any, will also be unaffected. That is, there is no economic consequence other than the gains and losses of direct utility, stemming from the monarchy itself, for the voters. Assume that for every voter the marginal utility of wealth is unaffected by the outcome. Then the only voters who would use Thompson insurance would be gamblers and those voters who disagree with the odds quoted for the insurance mechanism. All voters would use the VCG mechanism. Consequently the referendum would be similar to an ordinary vote, except that votes would be weighted by the strength of ordinary preferences and by voter wealth.

The third example, the choice between a military draft and an all-volunteer armed force, has elements of both the two previous cases. In practice the draft has involved underpaid involuntary servitude, often at considerable personal risk, for draft-eligible young men. In the United States their pay rates were apparently sometimes less than 60 percent of the pay that would have been necessary to recruit a volunteer force. Proponents of this system generally mentioned, among their reasons, that an all-volunteer force would be prohibitively costly, which is to say that these proponents wanted to hold down their own taxes by requiring the draftees to bear much of the financial cost. Assuming that potential draftees had proportional representation, or representation proportional to wealth, in the legislature, the proposed mechanism would rule out this kind of financing. The procedure for assessing approximate Lindahl taxes would result in full compensation for young men of military age, and the legislative mechanism would reveal that an all-volunteer force would be less costly than a draft. If a major group of voters nevertheless so strongly favored the principle of a draft that they would be willing to pay its extra cost, the referendum mechanism could choose a draft. To the extent that voters’ differences in hawkishness or dovishness were correlated with regional and other identified groups in the representation matrix, hawkish groups would pay higher taxes than the dovish ones to support the military forces.

These examples illustrate how a combination of mechanisms along the suggested lines can achieve efficient outcomes. They also illustrate how this approach can protect minority interests, or indeed majority interests, against abuse of the government’s power to tax and regulate.

4.7* THE PERSPECTIVE OF THE VOTING FAMILY

Having discussed the philosophical aspect of applying market simulation to government decisions, now let us consider two examples of how a referendum works out from the perspective of representative families in a small town. First we will consider an up-or-down vote on a budget item and then a budget item that can be approved anywhere in a range.

EXAMPLE(7)

Suppose that the firm currently supplying local police in the town of Freemonia, with 900 voting families, proposes an improvement in service using a new high-technology communication system. The Smith household’s tax assessment would rise by a one-time payment of $200 if the local referendum approves the proposal. The official odds of success published by the local Election Commission are .8 for approval and .2 for disapproval. The improved atmosphere of security, together with a small annual reduction in the homeowner’s insurance premium on the Smiths’ home, would raise their home’s property value by $210. The Smiths’ feeling of improved security against non-property crimes is worth an additional $90 one-time payment to them, of which $45 is the discounted present value of expected savings of health care costs and of avoided lost earnings. Thus, their expected gross gain is $300, of which $255 is material gain, which compares favorably with the tax assessment of $200. They would buy Thompson insurance for the $55 net material gain they stand to lose if the proposal is disapproved, at a cash premium of $11. In addition they would declare that disapproval would involve $45 of uninsured harm to them.

The vote turns out to be surprisingly close. The 500 families in favor insure for $15,000 of material harm and declare $5,040 of uninsured harm from failure of the proposal. The 400 families opposed insure for $10,000 of material loss and, many of them being concerned about a perceived threat to their civil liberties, declare $10,000 of uninsured harm from its adoption. The total dollar vote in favor is thus $20,040, compared to $20,000 opposed. A total of 29 voters, including one member of the Smith family, have declared $40 or more of uninsured gain from adoption of the proposal, and each of these owes some amount of VCG tax, which sums to a total VCG tax collection of $1,160. The total insurance premium collection is $11,000, the sum of $3,000 from those in favor and $8,000 from those opposed, so that after paying off the $10,000 insurance of those opposed and $800 of election cost, the Election Commission has a profit of $1,360.

The Smiths pay a VCG tax of $5. The Smiths have a perceived gain of $100 less their VCG tax of $5 and less their insurance premium of $11, for a net gain of $84. The constitutional provision for declared harms of these average sizes does not include the sophisticated adjustment for income effects of the technical annex, which would be negligible. It does require the procedure to conform to Article II, Subsection 10(g), refunding to each family a share of what the election profit would be without that family’s vote, however. The outcome without the Smiths would have been a loss of over $4,150, of which the share to the Smiths is an assessment of $4.62. For opponents and other non-pivotal voters the refund of election profit is about $1,335/899, or $1.48 each. The combination of refunds and assessments adds to a total net payment by the Election Commission of $1,155, leaving $205 to be placed in the election fund.

EXAMPLE(8)

The second example also concerns the town of Freemonia, together with a nearby group of riverfront properties over a five-mile stretch where the river is appreciably affected by pollution from the town and from the riverfront properties themselves. The town and riverfront area together have 1,000 voting families.

A private syndicate proposes a new sewer system serving the town and the affected group of riverfront properties, and also proposes industry restrictions (easements) governing industrial and commercial effluents into the river. The river in this area is suitable only for boating, although there are some carp in the river, considered to be inedible and of no interest to game fishermen. The proposal offers a range of pollution reductions from 50 percent to 90 percent, with project cost rising quadratically with the percent reduction. The townspeople benefit from savings in septic tank costs and from improved prospects for beneficial further development. The riverfront properties benefit from improved recreational values, with the possibility of upscale development. Using willingness-to-pay estimates from the local official legislatures, the Election Commission allocates the project cost function to households in the town and along the river. The estimated referendum outcome is a cleanup of 80 percent, with a 99 percent confidence range from 75.2 percent to 84.6 percent. The estimated probability of ‘no action’ is 0.1 percent.

The official election information bulletin from the Election Commission explains the costs and advantages of the river cleanup plan, with estimates of property value changes. It also explains the schedule of insurance premiums for purchasing insurance against the outcomes in the referendum range and explains how to report a family’s schedule of insurance purchases. Neighboring families get together to exchange views on the estimates and, when any of them knows an insurance agent, they bring her to the group to answer questions about the insurance procedure.

The Jones family lives by the river and expects that the project will change the character of the neighborhood, with a sharp increase in property values. The family’s allocated share of cost is $2,400 + $2x², where x is the percentage by which the project cleans up pollution. The Jones family expects their property to be worth $10,000 + $300x. In addition, because some unwanted neighbors, faced with the projected high tax assessments, will probably sell out and move if a high-cleanup project is approved, the Joneses have an added uninsurable $50x of willingness to pay for project approval when x is in the range from 80 to 85. The Jones’s prospective financial gain from the project is – $2,400 + $300x – $2x², and it works out that they would have a maximum net financial gain of $8,850 for a cleanup percentage of 75 percent. At higher percentages, their cost share rises faster than their property value. As just noted, they have uninsured willingness to pay of $50 per percentage point of additional cleanup, however, because (a) the Jones family plans to stay to enjoy improved recreation, which they value more highly than the expected market valuation, and (b) because higher cost increases their confidence that unwanted neighbors will move. (If the percentage of cleanup were to exceed about 85, the Jones family’s uninsured willingness to pay for additional cleanup would drop from $50 per point to smaller figures.)

The property value gain is insurable, so that they insure the $8,850 against no action for a premium of $8.85. Added cleanup would lower this gain, so that despite their desire for more cleanup the Jones family buys Thompson insurance against each higher outcome above 75 percent. For improvements in the referendum range, this reduction of financial gain ranges from about $1 for cleanup percentages from 75.2 to 76 to $170 for percentages near 84.6. The Jones family insures for these amounts by reporting its insurable loss schedule, namely $11,250 – 300x + 2x². (The incremental loss is less than $50 per percentage point throughout the range, so that the Jones family would prefer an outcome above the range.) The estimated probabilities of outcomes being a normal probability density over the stated range, the sum of the insurance premiums for the entire range works out at $55 plus small change. The family also reports their uninsured positive willingness to pay of $50x, which, contrary to their insurance, votes to enlarge the project.

The project wins approval at 82 percent cleanup, so that its insurable loss schedule gives an insurance payable to the Jones family of $98. Suppose that the typical marginal valuation schedule in Freemonia, like that of the Jones family, drops at the rate of about $4 for each percentage point of cleanup (given by the coefficient of x in the first derivative of the Jones family’s insurable loss schedule.) Calculating this information from the voters’ reported marginal valuation schedules, the Election Commission calculates the Jones family’s VCG tax as follows. At 82 percent cleanup these marginal valuations net of cost sum to zero, so that their sum without the uninsured declaration by the Jones family is – $50 per point, or an average of 5 cents per family by which cost exceeds willingness to pay. The typical family’s valuation schedule drops by this amount over a range of .0125 percentage points, so that without the Jones declaration of uninsured valuation the aggregate marginal valuation net of cost would sum to zero at a cleanup percentage .0125 less than 82. The Jones VCG tax is then approximately the area of a triangle with a base of .0125 and a height of $50, that is the amount (1/2)(.0125)($50)=$.31.

The Election Commission has a small profit from the insurance transactions and the VCG taxes collected from residents of Freemonia and of the riverfront area. The election commission disposes of the election profit in the same way as it did in the case of the police communication system.

Viewing these cases from a philosophical perspective, the constitutional plan has certain satisfying aspects that should have great appeal to thoughtful people at whatever income level. First, on budget issues and others involving a continuous scale of possible outcomes, every vote makes at least a small, if only symbolic, difference. If I am willing to pay to have a larger budget for police, education, or anything else, my vote will make it larger. Second, if I feel especially strongly about, say, education, and am willing to double my vote to increase its budget, my impact on actual budget size is doubled. No one would have a legitimate reason to feel excluded from the decision process. Reportedly, in present-day ‘democratic’ countries, they often have that feeling now.

In short, all in all it is easier to provide a reasoned defense of the proposed constitution than to provide one for the rule of one-person-one-vote, given the way that this rule works out in practice.

4.8* SOME FURTHER NOTES ON INCENTIVES

EXAMPLE (9)

The example of the high-tech police communication system for the town of Freemonia raises two issues that may need further clarification. One is the incentive effect of the residual profit that remains in the election fund after the Commission uses the constitutionally mandated procedure to refund its profit. The other is the possibility that the referendum has a negative profit. This outcome is most likely to occur if the balance of uninsured harms outweighs the insured harms, so as to require paying off more insured claims than the total premium revenue.

The net balance in the election fund, after refunds, creates an incentive problem unless it is transferred to the national election fund. In that case the effect of one family’s vote on that family’s refunds from all referenda becomes truly negligible if the national population numbers in the hundreds of thousands or higher. If the constitution does not provide for such a transfer, the incentive effect is hard to predict in a particular case, but nevertheless is inconsistent in strict logic with incentive compatibility.

If the Smith family were to increase its declared uninsured harm by one dollar, even though so doing would have no effect on the family’s primary refund (which is strictly unaffected by the family’s vote), it would affect the amount of final surplus or deficit in the election fund. The reason is that the declaration by the Smiths not only raises the winning balance in the referendum to $20041 in favor, but also raises the corresponding amount facing each other pivotal family ($20041 minus this family’s declaration of uninsured harm) for determining its VCG tax. That is, the amount of uninsured harm that such a family would need to declare to produce a tie is reduced by one dollar. For each pivotal family other than the Smiths, therefore, the effect is to reduce their VCG tax by one dollar. If, as assumed, there are 28 such families other than the Smiths, this effect reduces the surplus in the fund by $28. Further, there may be one or two families without whose vote (including insured harm) the outcome would go the other way by a margin of less than a dollar before the change in the Smith declaration. Such a family would be charged an assessment of about $4.62 before the change, but would instead receive a refund of about $1.48 after the dollar increase in the Smith declaration, because their vote no longer affects the outcome at all. If there were two such families, this effect would reduce the election fund surplus by a further $12.20, for a total reduction of about $40. Whether distributed in equal shares to all immediately or carried forward, the surplus affects the Smith financial position by 1/900 of the total. The effect of each dollar of Smith declaration on the eventual Smith financial position is $40/900 or $.0444, in effect increasing the Smith insurance premium and VCG tax by $100 times .0444, or $4.44. This effect is like a loading charge on their insurance, as well as an overcharge of the VCG tax. The Smiths would therefore have a corresponding incentive to reduce both their insured and uninsured declared harms.

These effects are the same for all families, whether pivotal or not, who favor the proposed police communication system. By contrast, an extra dollar of declared harm by a family who opposes the proposal would increase the VCG tax on pivotal families such as the Smiths, and would likewise have the reverse of the other above effects. Consequently such a family would have an incentive to overstate their insured and uninsured declared harms.

These effects on incentives apply whenever a household knows whether it will be on the winning or losing side, or when the household’s perceived probability of success of the proposal is different from .5. If a household has a Bayesian perspective, it would use this perceived probability, weigh its two incentives, and tilt its declarations accordingly. The result would be a small bias toward preserving the status quo. If the writers of the constitution wish to avoid this bias, they should consolidate all local election funds into the national fund.

The issue of losses on Thompson insurance is easier to deal with. Suppose that the election funds are consolidates as just suggested, so that the overall system is incentive compatible. In that case, the Smith family and every other family will in all cases declare a true combined total of insured and uninsured harms.

If a family disbelieves the official odds of the outcome, it will wish to falsify its insured harm but will offset this effect by an opposite change in its declaration of uninsured harm. Suppose, in an extreme case, that everyone is almost certain that the proposal will win, so that everyone considers it a good speculation to ‘insure against’ the proposal’s success. That includes the Smiths, who insure against it also. Like other supporters of the proposal, they will then declare an uninsured harm from the proposal’s failure of the amount of this insurance plus their true total harm from its failure, which for the Smiths is $100. The grand totals of declarations pro and con are shifted by equal amounts, and the proposal wins by $40 as before. The Election Commission has a big loss, and covers it with a special assessment, on average wiping out the various households’ profits on their speculation. The extent to which each household would push this strategy would be limited either by its risk aversion or by the constitutional limit on the household’s uninsured harm, whichever was more restrictive. (If their sense of certainty of the outcome pushes many of them to add substantially more insurance when their constitutional limit is binding, so that the extra insurance is unhedged, this exposure makes it likely that the proposal against whose success they are insuring will fail. The effect on the referendum outcome will teach them not to do that again. Most of the insurance account profit from this failure will be paid out as refunds to other families who didn’t overshoot.)

EXAMPLE(10)

The foregoing extreme example illustrates one incentive effect that complicates the overall picture. Another incentive effect arises when a family perceives an appreciable probability that its vote will be pivotal. In this case this family will have an incentive to insure more of its expected harm from a decision it opposes, and to declare less uninsured harm. This effect works the opposite way from the preceding one, and is similarly inconsequential.

The incentive to insure financial losses and (usually) not intangible losses comes from the diminishing marginal utility of income or wealth, that is, from risk aversion. If an intangible loss has no effect on the household’s marginal utility of income, the household will not buy insurance coverage for any part of the loss. However, if the household’s risk aversion is similar to that for U = log(Y), where Y is income, a modest subsidy to the insurance would make such coverage attractive.

An appreciable probability of paying the VCG tax provides such a subsidy. It works this way because adding a certain sum of insurance against the failure of a proposal has the same effect on the referendum outcome as adding that sum to a declaration of uninsured harm.

For simplicity, consider this point for a family who accept the official odds of the referendum outcome. Suppose that, instead of a .8 chance of success of the police communication system in the Freemonia referendum, the chance is given as .5, and that the Smith family believes that its chance of having a pivotal vote is .1. Suppose also that if its vote is pivotal, the conditional expected value of the family’s VCG tax is $22, so that its unconditional expected value is $2.20. Where otherwise the Smiths buy $55 of insurance and declare $45 of uninsured harm against the failure of the proposal, they would have the same expectation of putting the proposal over the top and at the same time would save the $2.20 of expected VCG tax by buying $100 of insurance against the proposal’s failure. The added $45 of insurance costs $22.50. If the Smith family’s income is $50,000, in round numbers this income rises by $50 if the proposal fails and rises by $5 if it succeeds (taking account of the tangible benefits of success), if the household buys the full $100 of insurance. Supposing that the Smith utility function can be accurately represented by U = ln(Y), over this range the marginal utility of income drops from 1/50005 to 1/50050. In that case the added insurance valued at this final margin is intrinsically worth $22.50 (50005/50050) = $22.48, or 2 cents less than it costs. The expected avoidance of the VCG tax is worth $2.20, making the substitution of insurance for a declaration of uninsured harm an attractive proposition.

The system thus provides various incentives to voting families to allocate their declarations between insurance and uninsured harm, depending on circumstances. In every case, however, the sum of insured and uninsured harm that each family declares is the true amount that the family is willing to pay to obtain the outcome it prefers. Typically those household opinions that disagree with official odds would be distributed randomly, so that the first, extreme case would not arise. Rumors, press campaigns, and so on could occasionally create an extreme case, however. In other cases (and perhaps in this case) many households would expect a close vote and would insure precisely their full willingness to pay. Our main point is that these cases make no difference to referendum outcomes.

4.9* INCONSISTENCIES

There are various ways that one declaration by a voter could contradict another declaration, although the main way would be through apparently inconsistent indications of risk aversion, perhaps because of differing kinds and amounts of publicity about two risks. For example, a voter might declare a willingness to pay $700 per year for a program that reduces the voter’s risk of death by .001 per year from heart disease, but a willingness to pay only $50 per year for a program that reduces the voter’s risk of death by .001 per year from a traffic accident. If queried by the election computer about this, the voter might respond that one can avoid traffic accidents by being careful. Suppose that the computer points out that almost half the victims of fatal traffic accidents were hit by a drunk driver or an errant truck, that is, under circumstances beyond the victim’s control, and that it also points out that life style changes can cut the risk of a heart attack by more than half. It could ask the voter, are these two cases as dissimilar as the voter’s declaration would indicate? (This information might have already been provided to voters in the Election Commission’s constitutionally mandated analysis of the advantages and disadvantages of the proposed programs.) The computer would give the voter some choices about ways to vote consistently, including an option to say that one’s family history makes one especially frightened of heart disease, as an explanation for the apparent inconsistency. That is, the voter could, in the end, insist, if this declaration did not later flatly contradict another.

If the voter’s declaration of harm from an unwanted outcome is incredibly high, the computer would point out the financial liability that this declaration could under some circumstances imply, beyond the voter’s means as declared on the family’s income and property tax returns. Even if the declaration could not bankrupt the voter, it might exceed the constitutionally set upper limit on declarations. In the case of an issue with a continuous set of possible outcomes, such as a budget, it would also point out to the voter the probable effect of the maximum permitted declaration on the outcome. It would ask the voter, is it worth that much? The voter would then decide what to do within the permitted limits.

These procedures would discourage careless or collusive responses. Although both kinds of responses could still occur, their effects would tend to be reduced to harmless proportions. The procedures would actively discourage collusive voting, because they would remind the voter of the incentive to defect from the collusive arrangement and to be truthful. They might also remind the voter that clusters of questionable votes could draw the attention of investigators looking into the possibility of unlawful collusion.

4.10* CONCLUSION

We have reviewed a number of possible arguments against our constitutional concept, and possible misunderstandings about it, and have found that on closer inspection it is robust against these arguments and misunderstandings. We elaborated on some of these points in the appendix to Chapter 2. Some of these arguments and misunderstandings, or the points of view that lead to them, have occupied whole books, and of course one could elaborate with book-length counter arguments. So long as our reasoning is clear and understandable, however, there is no need for that. We do elaborate further on technical issues in the remaining chapters and their appendices, to cover technical gaps whose solutions need to be gathered together in one place. On the broad philosophical issues involved in our proposal, we have completed our case.

A2.1* ALTRUISM AND PUBLIC-SPIRITEDNESS

In Chapter 1 we noted that gifts to charity, the act of voting, and the herd instinct fall outside the scope of an analysis that assumes only narrow, material-self-interested behavior, which I shall refer to as materialistic behavior. This entire work proceeds on the working premise that the narrow analysis that is traditional in economics is accurate enough for the purpose of constitutional design. The most direct challenge to this premise comes from Margolis (1982a, 1982b), who has argued that altruism invalidates the VCG mechanism. We argue in this appendix that it is in fact accurate enough for two reasons: (a) most of the non-materialistic behavior at issue has no effect on the analysis nor on the most effective constitutional design; and (b) where such behavior causes problems, the problems have no better solution than the proposed design.

To clarify these claims, we need to spell out the non-materialistic kinds of behavior that we can observe or that are suggested in the public choice literature. Consider first a married couple’s donations of income and of leisure time that fall outside their own material self-interest:

5. Participation in political or lobbying organizations, in public protests, etc.

The first item on this list usually receives a substantial fraction of a couple’s leisure time and income, on the order of one-third of the combined value of both, whereas the others usually receive five percent or less. Economists have therefore traditionally disregarded charities and political participation and have viewed the household utility function as an attribute of the entire nuclear family. That is, the welfare or specific consumption of each child was a variable parallel to the variables representing the parents’ consumption in the traditional utility function.

It turns out, however, that this traditional approach won’t do. When we look at long-term wealth accumulation, including the effects of taxation and government deficits, this approach has preposterous implications, because the children marry and their welfare then links the in-laws to each other, and then through an endless chain to everyone else (Bernheim, 1987; Bernheim and Bagwell, 1988). Recent work on this issue and on altruism in general has taken a different tack. Although a household’s gifts and bequests to children (and to others) may in part depend on the household’s concern for the welfare of the recipients, it is mainly controlled by the household’s satisfaction with the size of the gift. In technical terms, a household’s utility is entirely or partly a function of the size of the gift and only partly, if at all, a function of the perceived need, welfare, or the list of goods consumed by the recipient. Compared to the traditional way, this alternative way to characterize utility has notably distinct and different implications.

A2.2 PUBLIC GOODS

We also take account of public goods. For the present purpose we define a public good as a good in a government budget that wins approval in a referendum under the rules of Chapter 3. The motives for voter support can affect the analysis, but need not affect the definition of the good.

Of those goods and services that have been provided by modern governments, the following three kinds would very likely be continued by an efficient government using estimated Lindahl taxes:

Households would support Group 1 primarily, if not exclusively, for the material benefits its programs provide. Although the constitution mandates a minimal safety-net program, households paying estimated Lindahl taxes might support Group 2, over and above the constitutionally mandated transfers, for compassionate or public-spirited reasons. Households would typically have both directly self-interested and public-spirited motives for supporting group 3 programs. Programs in group 1 are likely to be comparative necessities, for which household willingness to pay would rise roughly in proportion to household income and wealth. Programs in group 2, and the public-spirited component of their willingness to pay for programs in group 3, could be comparative luxuries, so that willingness to pay could rise more than proportionately to household income and wealth, although that is not certain. Even if not, there would be few public goods for which household willingness to pay would fail to rise with income. Accurate Lindahl taxes would reduce but might not eliminate the sensitivity to income of the willingness to pay in excess of the household tax assessment, for successful programs. Consequently, as our analysis in Chapter 5 shows, there would typically be some uninsurable benefits of successful programs, so that the VCG mechanism would have a role in most referenda.

A2.3* ALTRUISM, PUBLIC GOODS, AND WILLINGNESS TO PAY

To illustrate the issues of most interest here, consider the ‘Smith’ utility function U = U(x, y, z), which the Smiths maximize subject to the budget constraint x + y = I(z), where we measure all variables in units of the numeraire, and where

A3 _ Extract